3 species force field

[hewuanhewuan]

Hello，
I want to train a reactive force field to simulate the reaction H2+O2=H2O2 on Pd. I tried to build four independent model, i.e. Pd/H2/O2 (three in one model), Pd bulk, Pd surface and H2/O2/H2O/H2O2 (four in one model), and use flare to train force field. It seems that the resulting force filed were terrible and produced some amazing and impossible substances. Are the four independent models I built reasonable? Or maybe there are some other mistakes I ignored？The otf parameters were pasted below.

otf: # On-the-fly training and MD
mode: fresh
md_engine: PyLAMMPS
md_kwargs:
command: "mpirun -n 64 lmp_intel_cpu_intelmpi"
specorder: [Pd, H, O]
dump_period: 1
pair_style: flare
fix: ["1 all nvt temp 300 300 $(100.0*dt)"]
keep_alive: False
initial_velocity: 300
dt: 0.001
number_of_steps: 50000
output_name: PdHO_otf
init_atoms: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59]
std_tolerance_factor: -0.1
max_atoms_added: -1
train_hyps: [0,20]
write_model: 4
update_style: threshold
update_threshold: 0.01
force_only: False

best,
hewuan